Struct optimization::StochasticGradientDescent

pub struct StochasticGradientDescent { /* fields omitted */ }

Provides stochastic Gradient Descent optimization.

Methods

impl StochasticGradientDescent

pub fn new() -> StochasticGradientDescent

Creates a new StochasticGradientDescent optimizer using the following defaults:

• step_width = 0.01
• mini_batch = 1
• max_iterations = 1000

The used random number generator is randomly seeded.

pub fn seed(&mut self, seed: u64) -> &mut Self

Seeds the random number generator using the supplied seed.

This is useful to create re-producable results.

pub fn max_iterations(&mut self, max_iterations: Option<u64>) -> &mut Self

Adjusts the number of maximally run iterations. A value of None instructs the optimizer to ignore the nubmer of iterations.

pub fn mini_batch(&mut self, mini_batch: usize) -> &mut Self

Adjusts the mini batch size, i.e., how many terms are considered in one step at most.

pub fn step_width(&mut self, step_width: f64) -> &mut Self

Adjusts the step size applied for each mini batch.

Trait Implementations

impl Default for StochasticGradientDescent

fn default() -> Self

Returns the "default value" for a type.

impl<F: Summation1> Minimizer<F> for StochasticGradientDescent

type Solution = Solution

Type of the solution the Minimizer returns.

fn minimize(&self, function: &F, initial_position: Vec<f64>) -> Solution

Performs the actual minimization and returns a solution that might be better than the initially provided one.